Detection of Small Defects on Photovoltaic Panel Surfaces Based on Improved Dn-YOLOv7 Algorithm
LIU Chengyi
DONG Xiaojie
LIU Sanjun
ZHOU Yunlei
Abstract：In order to mitigate the impact of image noise and tiny targets on detection of defects on photovoltaic panel surfaces,we proposed the improved de-noising you only look once version 7(Dn-YOLOv7)algorithm.Combined with de-noising convolutional neural network(DnCNN),a de-noise block(DnBlock)was proposed.The module used a loss function with a larger noise tolerance and utilized coordinates convolution(CoordConv)to integrate the noise channel,which enhanced noise reduction capability of the network.Meanwhile,the intersection over union(IoU)loss function was replaced with the normalized Gaussian Wasserstein distance(NWD)to improve model detection of small targets.The results indicated that the improved model had effective de-noise performance and higher detection precision.The mean average precision value reached 96.6%under no noise level.Under stronger Gaussian noise and impulse noise,the mean average precision reached 91.4%and 85.4%,respectively.The detection speed reached 78.0 frames/s.It has practical value in the detection of photovoltaic panel defects in aerial images.
Keywords：defects on photovoltaic panel surfacessmall targetdenoisingYOLOv7DnCNNCoordConvNWD
Publication Date：2024-06-20
Online Publishing Date：2025-08-15(First online date of this platform, not the publication date of the document)
Pages：7( 212-218 )